Integrates Core ML models into iOS apps with Swift APIs and MLModelConfiguration.
npx skills add https://github.com/FelixRauch/medat-figuren --skill coreml-felixrauch
Please help me install this Agent Skill. Skill: coreml Source: https://github.com/FelixRauch/medat-figuren/tree/main/.agents/skills/coreml Command: npx skills add https://github.com/FelixRauch/medat-figuren --skill coreml-felixrauch
This Skill solves the challenge of integrating Core ML models into iOS apps efficiently, reliably, and with good on-device performance so predictions run fast without blocking the UI.
.mlmodelc / .mlpackage using MLModelConfiguration, avoid repeated model instantiation, and support async loading.MLModelConfiguration and verify dispatch with MLComputePlan.MLFeatureProvider/MLDictionaryFeatureProvider, Vision integration, batch inference, and stateful inference using MLState (iOS 18+).CVPixelBuffer, use Vision to manage resizing/orientation, and work with MLTensor / MLMultiArray for numeric pipelines.Load a Core ML model from your app bundle with an MLModelConfiguration that selects appropriate compute units, then run an async prediction using the model’s typed input/output class (or MLDictionaryFeatureProvider) for your input data.
High-intent search queries and answers about installing and using this skill.
To run Core ML inference without blocking the UI, load your model using MLModelConfiguration and execute predictions asynchronously. This ensures fast on-device machine learning processing while keeping the main thread responsive.
Configure CPU, GPU, and Neural Engine dispatch by setting compute units in MLModelConfiguration. You can verify the actual hardware allocation and optimize performance using MLComputePlan profiling before running predictions.
The best way to handle image preprocessing for Core ML is using CVPixelBuffer alongside the Vision framework. Vision manages resizing and orientation, ensuring inputs correctly match your model's expected feature provider format.
Yes, you can run batched Core ML predictions and manage stateful inference using MLState on iOS 18+. This supports multi-model pipeline scenarios and handles typed inputs efficiently using auto-generated Swift classes.
Handle MLMultiArray and MLTensor data in numeric Core ML pipelines by passing them directly through MLFeatureProvider or MLDictionaryFeatureProvider. This ensures correct typed data handling for model inputs and outputs.
Avoiding repeated instantiation when loading .mlmodelc files prevents unnecessary memory overhead and latency. Reusing a single model instance with MLModelConfiguration enables efficient on-device loading and sustained prediction throughput.
femitzLoad and run Core ML models on-device with Swift APIs.
BorisserzLoad and run Core ML models on-device in Swift iOS apps.
mintuzIntegrate on-device AI models in iOS apps using Foundation Models and MLX Swift.
Roy-wonjiIntegrate Core ML models into iOS apps for on-device inference.